Deep Reinforcement Learning-based Data-Driven load Frequency Control for Microgrid

In order to overcome the frequency fluctuations caused by the uncertain random disturbances of microgrids, a data-driven load frequency control (DDB-LFC) method is proposed to achieve the multi-objective optimal control of control and economic performance. In addition, the soft actor-critic algorithm is adopted, and replaces the original LFC controller with an agent that can make independent decisions. It employs the soft actor-critic algorithm with strong robustness to realize the adaptive control of microgrids by reasonably setting the reward function of the agent. The simulation for the Zhuzhou isolated microgrid of Southern Grid (CSG) proves that the proposed algorithm can effectively reduce frequency deviation and generation cost.